Cooperative optical response of two-dimensional atomic arrays in solids Authors: Trevor Kling, Dong-yeop Na, Mahdi Hosseini Published: 2025 ABSTRACT AbstractWe develop a model describing long-range atom-atom interactions in a two-dimensional periodic or a-periodic lattice of optical centers considering spectral and spatial broadening effects. Using both analytical and numerical Green’s function techniques, we develop a mathematical framework to describe effective atom-atom interactions and collective behaviors in the presence of disorder. This framework is applicable to a broad range of quantum systems with arbitrary lattice geometries, including cold atoms, solid-state photonics, and superconducting platforms. The model can be used, for example, to scalably design quantum optical elements, e.g. a quantum lens, harnessing atomistic engineering (e.g. via ion implantation) of collective interactions in materials to enhance quantum properties at scale.Subject terms: Quantum optics, Quantum information, Single photons and quantum effects, Theoretical physics Introduction Recent advances in laser trapping of individual atoms1,2 have sparked interest within the scientific community to investigate the feasibility of constructing atomically thick optical elements. It has been proposed that a dilute 2D array of optically trapped atoms can be used as a mirror, utilizing the atom-atom interactions to enhance scattering3,4. An experimental realization was subsequently demonstrated based on a rectangular array of laser-cooled atoms5. Further investigation of the coherence of these arrays has revealed the formation of complex optical patterns in simple uniform arrays, even without a cavity6. In principle, these systems are not limited to only reflection, but can also be used to enable diffractive control of light. The continuously-driven atoms act as an array of apertures, allowing for the design of tailored diffraction patterns7,8. In the case of laser-trapped atoms, arbitrary control of atomic geometries can be achieved by holographic traps9 that enable realization of more complex collective interactions. Studying arrays of this kind in solids, rather than laser-cooled atoms in vacuum, offers the advantage of scalability, with atom positions controlled by focused ion implantation below the diffraction limit10 and with single-ion resolution10. Previously, we developed theoretical and experimental models to study 1D arrays of rare earth ions in solid-state photonic devices11–13. In this article, we develop a theory to model the interaction of light with 2D array of atoms (or quantum optical centers) of arbitrary geometries within a solid-state host, accounting for statistical variations in atomic frequency, atom number, and atom locations. With recent advances in ion implantation10,14, material interfaces15, and solid-state spin qubit16,17 developments, our model allows for design of efficient and large-scale collective effects among millions of optical centers in solids for applications including quantum storage18, entanglement creation19 and superradiance interaction20–22. State-of-the-art implantation techniques allow localization of implanted ions within tens of nanometers of a desired point14,23, and prior work from our group has shown the emergence of long-range collective behavior from implanted 1D arrays of rare-earth ions21. However, while solid-state systems can support any arbitrary geometry, the implanted ions have their resonant frequencies shifted due to fluctuations in the local crystal field. The implanted ions form a spectral distribution due to this inhomogeneous broadening, which reduces the effective number of atoms at any chosen frequency within the broadened linewidth. By doping or implanting a large number of these atoms, a similarly dilute array of resonant ions can be produced in a solid-state host. In this paper, we consider 2D geometries of emitters within a solid-state host and theoretically study collective behavior emerging from long-range atom-atom coupling. We devise a model and develop a mathematical and numerical framework to capture the effects of broadening as, for example, seen in solid-state materials hosting ensemble of quantum centers, e.g. rare-earth ions. The model allows exploration of arbitrary atomic geometries with realistic disorder to optimize certain collective properties. Scattering dynamics Lattice modes have been the subject of extensive prior investigation for their collective properties. Theoretical works describing these collective behaviors have evaluated the structure of the resulting superradiant and subradiant modes for various uniform geometries in one, two, and three dimensions24–26. These studies focus on the case of fully ordered atomic systems, as may be achieved by laser trapping atoms. Lattices built based on identical atoms have been shown to be robust against small variations in their lattice spacing3, and as such prior models typically disregard sources of disorder in the system. However, for a lattice built within a solid-state substrate, intrinsic and fabrication disorder become significant factors. Understanding the dynamic and static behavior of atomic arrays inside solids is becoming increasingly important given recent advances in scalable solid-state quantum material engineering. Atom-atom interactions have been evaluated theoretically and numerically in the case of atomic gases27,28. Here, the atoms are distributed within a cloud given by a particular probability distribution. Importantly, these systems exhibit strong deviations from the mean-field behavior, except in the case of large inhomogeneous broadening29. In this paper, we develop a model to describe light-atom interactions in an engineered array of solid-state spins or quantum centers. This differs from prior studies, as the theory captures lattice phenomena as seen in the trapped atom case, but includes the limitations of disorder as seen in the distribution of atoms in a cloud. This theory provides a bridge between between the two aforementioned topics; the low-disorder limit the system behaves similarly to a system of trapped atoms, and when the disorder becomes larger than the lattice spacing, the behavior should be similar to that of a randomly distributed gas. Our model can be applied to realistic systems of trapped atoms and solid-state atomic arrays to optimize atomic geometries and enhance specific properties, even in the presence of frequency and positional broadening. For an embedded system, arbitrary geometries of quantum centers can be realized, for example, by ion implantation techniques. Here, we design a spiral atomic array to produce an “atomic lens,” where the light scattered by the atoms is focused to a point in the far-field. This may allow for more efficient detection of scattering from the atoms, while the single-arm spiral geometry still allows all atoms to couple to common modes propagating along the arm of the spiral. Other diffractive grating schemes, such as a Fresnel zone plate, may not ensure coherent behavior between spatially separated zones. For optical applications, we employ a spiral geometry similar to those used in prior studies of hollow beams30,31, as shown in Fig. 1. This semi-1D geometry has the benefit of focusing scattered light in the far-field, while maintaining a continuous profile such that all atoms may interact with a common planar mode. Tailoring the light to produce such a distribution has applications in atomic trapping. The geometry of the spiral is determined by the desired focal length F and the wavelength of the light . We consider the case of a spiral with a single arm, producing a hollow beam with topological charge . Eq. 1 provides the parametric equation for a continuous 1D spiral; to translate this behavior to an array of atoms, the spiral is given a finite radial width and atoms are placed with uniform spacing inside the region. To facilitate focusing, we include a central opaque region by restricting . Locally, we retain the square lattice distribution as in a complete plane. The atoms are located at positions , determined as the points on a uniform square grid of spacing a that fall within a specified length from the radius of the spiral arm in Eq. 1. Generally, this could be viewed as “cutting” the opaque regions of a spiral Fresnel zone plate out of a complete square grid. We consider exciting the atomic ensemble with linearly polarized light, so maintaining orthogonal symmetry in the local geometry better maintains the polarization of the reflection and increases the cooperative effects. (a) Vertical view: atoms are doped into the substrate in a spiral structure following Eq. 1. Each implantation region is separated by a distance a in a square lattice, along a spiral determined by Eq. 1. (b) Horizontal View: the atoms implanted a distance d into the substrate with permittivity are exposed to a uniformly polarized incident field . The reflected field is then focused a distance F above the atoms, in the free-space region with permittivity . (c) 3D view: the phase relationship of the light reflected by a radially-varying set of atoms results in the formation of a hollow beam at the focal point. To probe the optical response of the system, the atoms are excited by a monochromatic, normally-incident Gaussian beam with uniform polarization in the x direction. To investigate the behavior of the ensemble under these conditions, we model the atoms as classical dipoles in the weak-excitation regime, assuming a linear isotropic polarizability3,32–34. In this work, we consider a broadened ensemble of atoms such that the polarizability of each atom is dependent on their resonant frequency. To describe the behavior of these atoms, we begin with the theory corresponding to an ideal lattice of atoms. When the system is exposed to an external field, the atoms are driven by both the initial incident field as well as a scattered field from all nearby atoms . A description of how the incident and scattered field are connected is provided in Sec. S1 (Supplementary Materials). Depending on the interatomic spacing, the sum of these fields can be either constructive or destructive, resulting in enhanced or suppressed polarization of the total ensemble. The induced dipole moments can be found by solving the system of linear equations given in Eq. 2. Here, k is the angular wavevector of the scattered light, is the permittivity of the medium, and and are the position and polarizability of each atom. , the initial dipole moment, is set by the incident field as . A detailed description of the Green’s function is provided in Sec. S2 (Supplementary Materials), for all cases described in this work. Each atom is assumed to have a uniform homogeneous linewidth regardless of its broadened frequency. To properly include inhomogeneous broadening, the polarizability must allow for variation in both the frequency of the pump and the resonant frequency of the interacting atom. We define the polarizability of an individual atom in Eq. 3, where is the frequency of the pump, is the frequency of the interacting atom with corresponding wavelength and wave vector , is the refractive index of the material, and is the total decay rate35. Embedded array When the scattering atoms are implanted into a solid-state medium, the behavior of the system is perturbed in three main ways. Firstly, the presence of a dielectric interface alters the propagation of the scattered light. This manifests in two effects; a modification to the Green’s function corresponding to the reflected light, and a position-dependent phase shift for the transmitted light. Secondly, atoms implanted into the solid experience inhomogeneous broadening arising from local defects in the host. This results in a randomization of the phase of the scattering, discussed in greater detail in Section IV. Finally, implantation procedures have a finite precision in terms of the doping location. This introduces positional disorder of the implanted atoms, characterized by a finite uncertainty in the location of the scatterers. Atoms are implanted in a circular region with radius , typically on the order of a small fraction of the resonant wavelength. These implanted atoms experience inhomogeneous broadening due to local variations from the solid-state host material; as a result, the probability that a given ion has resonant frequency is follows a Gaussian inhomogeneous distribution. To account for the inhomogeneous broadening, we consider the effective polarizability of one “site” within the lattice, rather than a singular atom. Each site contains N atoms with inhomogeneous broadening. For a pump field at a fixed frequency , only a fraction of the implanted atoms will be excited. To account for this, we define an effective polarizability of each implantation site based on an average of the contribution of each implanted atom. The distribution of these atoms is a Gaussian, and the average site polarizability can be found in Eq. 4. Here, we average over a range of to incorporate all atoms with sufficiently strong interaction with the pump field. The term here is a binomial distribution included to model the effects of radiative quenching. If multiple atoms with similar frequencies are embedded within the same site, their large mutual interaction results in radiative quenching. In typical solid-state systems, e.g. rare-earth ions or defect centers in solids, the homogeneous linewidth is typically much less than the inhomogeneous linewidth. Using this assumption, the effective polarizability of the site is given by 5, which provides the polarizability of a singular site for any pump frequency . Here, for the inhomogeneous linewidth , and is the central frequency of the inhomogeneous distribution. is a constant corresponding to an integral of the Lorentzian distribution with width , integrated over a range . A derivation of this form and further definitions are provided in Sec. S3 (Supplementary Materials). This equation can be understood as the resonant response of an effective atom, , multiplied by an effective atom number . For the following numerical analysis of these systems, the atoms are assumed to be placed in a half-space of material with a depth d. The scattered field is then given by Eq. 6, The lattice mode of interest is formed by local interactions between atoms within an arm of the spiral, and direct long-range coupling contributes little to the observed effect. Instead, collective interaction is facilitated by the lattice mode supported by the sub-wavelength distribution of atoms. We confirm the existence of a focal region in spite of the dielectric interface via finite-difference time-domain (FDTD) simulations allowing reflections. For details of the FDTD simulation, please refer to Supplementary Materials Figures S1-S3 and Table S1). For the numerical parameters of the solid, we opt to match our system to a common quantum-optical system of rare-earth atoms doped in crystal at cryogenic temperatures surrounded by vacuum, such as thulium-doped Lithium Niobate or Erbium-doped Silicon-28. Our choice is motivated by the fact that the production of the kinds of devices simulated here can be achieved using modern nanofabrication and implantation techniques on commercially available wafers. The rare-earth atoms considered here typically demonstrate similar spectral behavior in bulk crystals and also integrated photonics36–44. The semiclassical dipole methods employed correspond to the weakly-excited regime of the array, where spectral effects due to driving of atoms, such as spectral hole burning, is expected to be negligible. We consider a system with ensemble-averaged implantation sites, including effects such as spectral diffusion as modifications to the linewidth 37. The numerical values of the homogeneous and inhomogeneous linewidth used in this work are taken from prior experimental work Ref37.. The effect of temperature inducing decoherence is a known effect but cannot be predicted or computed precisely unless we restrict ourselves to a specific material and host. Here we consider rare earth ions in solids. The interaction with phonons can happen through direct, Raman, or Orbach processes, all of which contribute to the total relaxation rate45. This effect has been studied in-depth and in this work, we consider an effective and realistic relaxation rate in our simulation. It has been shown that at low enough temperatures and high enough magnetic fields, coherence time of some rare earth ions can reach minutes or hours, and the inhomogeneous broadening is no longer limited by temperature. Of great interest in these systems is the effect of interaction between neighboring sites. A dilute array interacts with both the incident field and the field scattered by all neighboring atoms, constituted by the second term in Eq. 2. Each of these neighboring sites then contributes an interaction term that constitutes a phase shift and dissipation, dependent on the spatial relationship between the interacting sites. To model the collective behavior of an atomic array, we model atom-atom coupling using a simple Hamiltonian given by Eq. 7. Here, n and m index implantation sites and j and l index dipole components. Assuming that our atoms are isotropic, the dipole components can be given by , and the simplified Hamiltonian is given by Eq. 8. The collective behavior of the system can be evaluated by finding the eigenvalues and eigenstates of this Hamiltonian; highly collective states will correspond to highly mixed eigenstates, and their eigenvalue will be the corresponding resonant frequency and decay rate modification of the collective mode. The coefficient of the second term in this Hamiltonian can be reinterpreted as the radiative frequency and linewidth shift of the corresponding atom induced by the atom-atom interaction. Summing over these coefficients gives a measure of the cooperative behavior of the ensemble. To better interpret the effect of atom-atom coupling, we can define the real and imaginary parts of the coupling between atoms n and m as Eq. 9 For every pair of atoms, the dipoles may couple along any of the three spatial axes, corresponding to the lower indices j and l. It is convenient to express this as a tensor of the spatial components, . For a uniformly excited array, the total collective phase shift and linewidth shift can be given by Eq. 10. To demonstrate the effect of this detuning on the scattered field, we can consider an idealized 2D square array, with uniform spacing a between all atoms without variation. In this scenario, the collective phase shift and collective linewidth shift are uniform for all atoms, as a consequence of the uniformity of the Green’s Function. The 2D translational symmetry allows for closed-form analytical expressions for the ensemble-wide polarizability of the atoms3, given by Equation 11. In this form, the dipole moment of the interacting ensemble under uniform excitation from an incident field is given by . This dipole moment produces the scattered field, as shown in Eq. 6. For a system excited with a uniform x-polarized incident field, only the xx component of this coupling is significant; the off-diagonal xy coupling is more than an order of magnitude smaller than the on-diagonal components. Therefore, under these conditions one may identify and as the principal values dictating the collective behavior of the system. When considering non-uniform geometries or platforms with significant spatial uncertainty for the atoms, the analytical methods used to reach Eq. 11 are no longer applicable. However, the nonuniform magnitude of the Green’s function for varying interaction differences implies that the local interactions between atoms dominate the collective behavior. The simulations in this work support this assertion, as the spiral arrays display a localized maximum dependent on the collective effect. For comparison, Fig. 2 shows the calculated calculated scattering intensity and lattice-induced detuning for free-space square and spiral lattices. Here, each lattice site hosts exactly one atom at the resonant frequency. In both cases, the reflected intensity I achieves a maximum around , corresponding to zero lattice-induced detuning. The square lattice demonstrates a secondary local maximum around , which is not present in the spiral case. For an infinite square lattice, the symmetry of the lattice enhances scattering from the lattice proportional to 3. This compensates for the reduced polarizability of the ensemble at small lattice spacings, resulting in the second observed peak. Maximum reflected intensity () and ensemble average lattice mode detuning for an ensemble of atoms. (a) A square lattice of atoms with spacing a. The maximum intensity is taken from a region sufficiently far from the plane to eliminate local effects, and is normalized to the highest intensity observed while varying the lattice spacing. (b) A 2000-atom spiral with a m focal length. The maximum intensity is taken from the focal region. Numerical simulations were performed with parameters . Here, the condition corresponds to the frequency of the incident field being on-resonance with the cooperative resonance produced by the array of atoms. We note here that this does not imply that the atom-atom interaction vanishes, but rather that the magnitude of the detuning induced by coherent sum of all atom-atom interactions is minimized. The single-atom parameter is averaged over the complete ensemble, and thus represents a measure of the collective behavior of the lattice. While two resonances are observed in these systems, the physics of each resonance is distinct. For small lattice spacing, the high atom-atom interaction leads to superradiant emission, with the cooperative linewidth shift diverging as a goes to zero. The secondary resonance at larger atomic spacing exhibits subradiant behavior, with . To investigate the focusing effect of electromagnetic waves from a spiral-shaped array of atoms thinly embedded in a dielectric medium, numerical simulations of the scattered electromagnetic field were performed. Simulations of the scattered field intensity were carried out on the Quest High-Performance Computing cluster at Northwestern University, while a spatially-resolved simulation of the focusing behavior with a dielectric interface were completed on a workstation at Pohang University of Science and Technology (POSTECH), South Korea. To capture the effects of an interface on the collective lattice mode, a three-dimensional FDTD algorithm was employed. A detailed description of the numerical method is provided in Sec. S4 (Supplementary Materials). To verify that the diffractive focusing is preserved in the presence of a dielectric interface, Fig. 3 depicts x, y, and z components of the electric field obtained from the three-dimensional FDTD simulation at time-step index on the yz plane. Simulations only involving the focal plane of the structure can be managed efficiently by using the far-field forms of the scattered field. However, simulations covering all of space must use the near-field equations, which for the approach employed in 6 involves computing a large number of numerical integrals as well as dealing with the singular nature of the Green’s function. FDTD simulations bypass these constraints and can compute the field distribution and Green’s function for an arbitrary distribution of scatters and dielectrics, though they do not include atom-atom interactions. Here, we demonstrate the focusing behavior of the adjusted site geometry. Since the dipole moment of the atoms is aligned along the x-axis, the x component of the electric field is strongly formed, and it can be observed that a strong focusing spot occurs approximately 20 m away from the interface of the dielectric in the vacuum region as expected. x, y, z components of electric fields obtained from the three-dimensional finite-difference time-domain (FDTD) simulations at time-step index . The spatial grid size was taken to be , and each temporal time step spanned a time . Atoms inside the solid host are excited by a x-polarized Gaussian beam, and the formation of a hollow beam is observed at a distance F from the array. When considering dipoles implanted shallowly within a dielectric medium, one additionally must adjust the spiral parameters to account for refraction from the interface. Specifically, consider a transition between a region of refractive index and , with . Employing standard ray optics, Eq. 12 gives the new path length taken for a source implanted at a distance d beneath the dielectric surface and at radial distance R from the central axis to reach a point elevated a distance F above the interface. The first term here is identical to the path length for the free-space spiral, while the second term gives the adjustment to the path length due to the propagation inside the medium. This path is subject to Eq. 13, the phase-matching condition for constructive interference. This equation ensures that, at the focal point F, the light from emitters separated by an angular distance of will have a relative phase shift of , producing a hollow beam. Solving this equation numerically for produces a spiral with varying curvature throughout its angular coordinate, adjusting for the variance in refraction due to changing incident angle. Solving Eq. 14 gives the atom radius for an implantation depth d, adjusting for refraction due to the dielectric interface. This equation adjusts the atom position to match the surface emitter location to the bulk emitter location, including both angular and optical length considerations. For a typical Fresnel-type diffraction grating, one would expect to observe a series of secondary foci at each odd harmonic of the original focus, 46. However, the adjustments required to optimize the primary focus suppress these additional foci as the radial modifications are only constructive for the primary focus. Due to limitations of the FDTD simulations, these plots disregard the atom-atom coupling considered in the other calculations of focal behavior. Even though the FDTD simulation does not account for atom-atom coupling, it is useful to calculate classical effects at the interfaces. The FDTD simulations allow for the retrieval of the electromagnetic Green’s function, which is used in all other simulations to model atom-atom interactions. Extracting these factors from the simulation allows validation of the computational methods used to calculate the atom-atom interactions and ensure numerical convergence. Varying the spacing of the underlying lattice a allows for control of the lattice mode, as shown in Fig. 4.a. Here, the pump frequency is set to the central frequency of the inhomogeneous distribution. The maximum reflected intensity is observed in the vicinity of . This corresponds to a maximization of the collective behavior of the array, indicated by the small reduction in the collective decay rate. The ensemble-averaged atomic polarization, discussed in the following section, provides a measure of this collective interaction. (a) Reflection from an embedded spiral lattice with inhomogeneous broadening and surface reflections versus lattice spacing. The blue curve indicates the reflected field intensity at the focal point of normalized to the reflection from a spiral with atoms distributed randomly within its arms, . The orange curve correspond to the cooperative linewidth shift of the atoms, averaged over the ensemble. Numerical simulations were performed with parameters . The pump frequency is taken to be the same as the atomic frequency. Effects due to nonradiative decay were ignored, taking . (b) Frequency broadening for a free-space linear spiral with atomic spacing . Scattering is normalized to a system with equal density of atoms but no lattice structure. (c) Purely positional broadening for an embedded spiral with spacing and . Centers with more than one atom implanted were ignored to avoid changes in the effective number of scatters distorting the trend in the intensity from the lattice mode. Note that for large site radii, the scattering intensity decreases linearly with radius due to the loss of spatial definition of the diffractive structure (in this case, the spiral). This effect becomes pronounced around . The inset provides a visual representation of the radial variation. Broadening analysis Inhomogeneous linewidth The effect of the inhomogeneous linewidth on the collective behavior is examined in Fig. 4.b. Here an ideal array of atoms is assigned a varying inhomogeneous linewidth, and the reflected intensity and ensemble-averaged dipole moment are simulated. The number of dopants is varied along with the inhomogeneous linewidth, such that the single-site polarizability is maximized throughout. The system can be categorized into three main regions; one where the inhomogeneous linewidth is much less than the homogeneous linewidth, one where the inhomogeneous linewidth is much greater than the homogeneous linewidth, and a transitional region where the two are similar in magnitude. When the inhomogeneous linewidth is much less than the homogeneous linewidth, the system behaves as an approximately uniform 2D layer of atoms. The broadening here has little effect on the lattice mode, and the behavior of the system is consistent with prior theoretical analysis such as3. Here, the probability distribution in Eq. 4 approaches a delta distribution as the inhomogeneous linewidth is a fraction of the atomic linewidth. The behavior of such systems is consistent with ensembles of laser-trapped atoms. When the inhomogeneous linewidth greatly exceeds the atomic linewidth, the system enters into a uniformly-distributed case. The polarizability of atoms in this scenario was discussed previously in Eq. 5, where the Gaussian distribution modeling the inhomogeneous broadening is approximately flat over the integration region. This relation is maintained regardless of choice of pump frequency; tuning the pump within the inhomogeneous linewidth can change the effective atom number, but the uniformity of the spectral distribution is generally unchanged. This case covers the majority of typically-employed doped optical materials. Defects which affect the degree of inhomogeneous broadening also impact the homogeneous linewidth; as a result, systems with low inhomogeneous broadening typically also have a comparatively narrow homogeneous linewidth. We note here that the exact width of the inhomogeneous broadening is largely unimportant with regards to the lattice mode, so long as the inhomogeneous width is more than an order of magnitude greater than the homogeneous linewidth. However, the larger inhomogeneous linewidths require larger dopant concentration in order to achieve equal average population. As a result, in practice one would expect that system with greater broadening would experience reduced spin coherence and other detrimental effects from crystal damage or defects at high densities. When the inhomogeneous linewidth is similar to the homogeneous linewidth, the system enters into a distinct behavior regime where the lattice mode has a strong dependence on the pump frequency. Here, a monochromatic incident field can couple to a nonuniform distribution of atoms. When the pump frequency matches the central frequency of the inhomogeneous distribution, the system exhibits a symmetric response with an approximately equal number of positively detuned and negatively detuned atoms excited. However, shifting the pump frequency far from resonance produces an asymmetric excitation distribution, favoring atoms detuned in one particular direction. Spatial variance When the atoms are implanted into a material, the precise atomic positions exhibit spatial variation depending on the method of doping and annealing. This positional variation both influences the strength of coupling between local atoms, as well as shifting the phase of the scattered field. Increasing this variation results in a incoherent phase background which suppresses the coherent behavior of the lattice, as shown in Fig. 4.c. However, the effects of the spatial variation and inhomogeneous broadening are not necessarily additive; frequency shifts and spatial shifts can result in opposing phase effects, effectively restoring the ideal lattice behavior. To analyze the effect of spatial broadening, we consider a simple system of two implanted atoms separated by a distance a with a uniform circular implantation region of radius R. Assume, for simplicity, that the displacement is along the x-axis. For any pair of displacements in this region , there exists a similar pair with equal radius but opposite change in angle found by mirroring the points across the x-axis. Thus, when the effect of this broadening is averaged over a sufficiently large ensemble, the ensemble-wide change in angle between pairs of points averages to zero. However, this is not true for the radial displacement. Consider a semicircular region created by the intersection of a disk centered at the second atom with radius a and the uniform region of implantation centered at the first atom. Any point on the boundary of this region maintains the radial distance between these points, while any point on the interior reduces it and any point on the exterior extends it. By inspection, it is clear that there are more points outside the intersection than inside it, and therefore the effect of the uniform distribution is simply an increase to the effective distance between points. Additionally, this increase in distance is always along the axis of interaction, due to the intersection region being symmetric about its axis. This applies to both regions; thus, the number of pairs of points where the shifted radius is increased exceeds the number of pairs of points where the radius is reduced. In Fig. 4.c including spatial broadening shifts the peak observed at a spacing of to a reduced spacing, as the effective separation grows. However, this effect is nonlinear, as the new effective radius depends on both the size of the implantation region as well as the schematic of this behavior is provided in the inset of 4.c. Dopant concentration The final source of error in these array systems is the number of atoms per region. When multiple atoms near the same resonant frequency are doped within the same region, they have a much stronger interaction than the engineered atom-atom interaction in the array. As a result, the phase and amplitude of their collective polarization is strongly affected; for atoms much less than a wavelength apart, this results in radiative quenching. For atoms farther apart, the dipoles orient along the common axis. This produces scattering with highly incoherent phase from these atoms, reducing both the collective effect of the array and the interference pattern in the focal plane. Discussion The above analysis is applicable to a wide range of quantum center arrays. For example, rare-earth ions embedded in a solid-state host such as a lithium niobate slab can be engineered into arrays with arbitrary geometries. As a multi-functional photonic material, lithium niobate has broad applications in classical and quantum photonics47. Additionally, advances in nanofabrication techniques have resulted in commercially available lithium-niobate-on-insulator (LNOI) wafers, which allow for nanostructures with the aforementioned properties48,49.. The integration of active quantum centers such as rare earth ions with LNOI creates additional opportunities for quantum information control and processing11,48,50,51. Creating emitter arrays in solids faces several key challenges. The foremost is achieving precise emitter localization, especially for ions with shorter resonant wavelengths that demand sub-wavelength implantation accuracy. Controlling decoherence and inhomogeneous broadening remains difficult, particularly for emitters near surfaces where implantation-induced defects degrade performance. While post-implantation annealing can reduce such defects, it risks atomic diffusion and loss of spatial precision. A promising path forward is the use of isotopically purified Si doped with rare-earth ions, which offers long coherence times, low broadening, and compatibility with nanofabrication techniques that can define array geometries without relying on implantation15,43,52. We note that for the simulations in this paper, choice of material or quantum centers result in minimal qualitative change to the observed behavior. However, the reduced inhomogeneous linewidth requires a smaller dosage of ions to resolve the same average ions per site. As a result, we anticipate that certain experimental properties, such as coherence time, will be improved for a silicon-substrate array compared to LNOI. In these simulations, we observe a clear lattice resonance from the embedded arrays, even with a large inhomogeneous linewidth. However, the phase randomization from the change in frequency of the scatterers results in a weaker lattice resonance compared to a free-space case; the lattice mode-induced detuning in the embedded case only achieves 10% of the effect observed in the zero-broadening designs. This is caused by the incoherent phase background from the inhomogeneous broadening applied alongside the coherent spatial phase. We note here that we have considered a system with significantly large inhomogeneous broadening; an increased effect may be observed using state-of-the-art low-broadening systems such as isotopically purified silicon or silicon carbide. At large broadening, one can consider other geometries to improve cooperative behavior. While the collective lattice modes are still observed in the embedded regime, we suggest that a non-uniform spatial distribution may produce even better results. Prior simulations of free-space systems has shown that the phase shift of a fully detuned ensemble can be compensated for by modifying the lattice spacing3. As such, a sufficiently complex geometry may improve the average behavior by creating an array such that a certain number of nuclei achieve optimal behavior regardless of the single-atom frequencies. We note here that the focusing effects from a diffractive lens does not require uniform reflection from all points; classical disorganized lenses have been employed in the form of photon sieves for the purposes of broadband performance and control of higher-order modes7,53. Rather, we anticipate a geometry designed with highly nonuniform spacings such that each major region of the spiral would contain some number of atoms with near-ideal couplings. One avenue to search for such geometries could be the implementation of fractals. The atom-atom coupling can be interpreted as a form of random walk; such Brownian motion is well-known to produce fractal structures via diffusion-limited aggregation. A fractal structure may provide the kind of dense phase relationship required to improve performance of these systems in realistic conditions. Alternatively, one could search the geometric space via machine learning. Studying properties like the atomic polarization and scattered intensity as optimization targets, one could search for arbitrary 2D geometries that maximize the desired behavior. Such approaches have been employed to create arbitrary diffraction patterns in traditional optics54 and to optimize metasurfaces for optics applications55. Such systems could be used to construct atomically thick lenses and other diffractive optics56. With focusing provided by the geometry itself, such systems would have promising applications for squeezed light generation and other nonlinear processes. Unlike a refractive lens, the focusing here is produced by the phase relationship between the different atoms participating in the scattering. Thus, under conditions where the ensemble remains coherent throughout its lifetime, one would expect to observe the same focusing effects from the resulting spontaneous emission. To realize such atomically thick lenses, a focused ion beam10,14,57 or a nanoscale implantation mask23 can be used to implant isotopically purified emitters (e.g. rare earth ions) with spatial resolution of 10 nm or less and atom resolution as low as a single atom. The implantation energy can be low to avoid crystal damage, while depth can be controlled by additional material growth after implantation58. Localizing ions or defects via implantation in hosts with exceptionally low inhomogeneous broadening such as isotopically pure Si15,43 or silicon carbide59 opens doors to atomistic design of quantum lenses, and other diffractive optics exhibiting large-scale and scalable collective behavior. Conclusion In this paper, we developed a theory to investigate collective interaction of light with an arbitrary array of atoms in a dielectric medium. Unlike atomic arrays studied in free-space geometries3,25,60,61, arrays embedded in solid hosts exhibit additional complexities such as inhomogeneous broadening, positional disorder, dielectric interfaces, and lattice coupling. These effects necessitate a distinct theoretical framework for describing light–atom interactions in two-dimensional atomic arrays within solids. This is particularly relevant given recent advances in atomic-scale device engineering21,62, which enable the fabrication of arbitrarily large, on-chip atomic arrays without the need for laser cooling. In this work, we develop a theoretical framework that incorporates various broadening mechanisms and interface effects in the collective light–atom interactions of solid-state arrays. Our analytical and numerical methods capture realistic broadening in solid hosts and are applicable to a wide range of quantum systems, including lattices of rare-earth ions in photonic materials. Using a Green’s function formalism and numerical simulations, we calculate the atomic polarizability, emission intensity, and lifetime as functions of broadening parameters. We demonstrate that, despite the presence of broadening and disorder, cooperative effects persist in solid-state atomic arrays under realistic experimental conditions. We also discuss how materials can be engineered in a top-down approach to harness collective behavior in atomic arrays and build atomic-thick optical elements for quantum and classical applications. Supplementary Information Below is the link to the electronic supplementary material. Supplementary material 1 Author contributions T Kling has performed analytical quantum calculation, performed quantum simulations, and prepared the first draft of the manuscript. DY Na has performed semiclassical simulations and finite element simulation. M Hosseini has supervised the work. All authors reviewed and revised the manuscript. Funding TK and MH would like to acknowledge support from National Science Foundation Award No. 2410054, and DN would like to acknowledge support from the National Research Foundation funded by the Korean Government (MIST), under Grant No. 2023-00213915. Data availability The datasets used and/or analysed during the current study available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Dong-yeop Na, Email: dyna22@postech.ac.kr. Mahdi Hosseini, Email: mh@northwestern.edu. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-31153-4. References Associated Data Supplementary Materials Supplementary material 1 Data Availability Statement The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.